A prompt injection flaw in Google Gemini's voice assistant let attackers hide malicious commands in notifications, enabling social engineering and more.
Malicious Notifications Could Trick Google Gemini Users
Résumé + lien vers la source. Sans paywall, sans tracking.